It is possible that takeInhibitor() throws (see log below). Therefore,
ensure that leaveAction is executed by using a try/finally block.

Jan 19 09:01:56 host sandmann[1574]: Invoking /etc/sandmann/activity-check.d/bitcoin-lag
Jan 19 09:01:56 host sandmann[1574]: Invoking /etc/sandmann/activity-check.d/bcache1-dirty
Jan 19 09:00:46 host sandmann[1574]: Received sleep signal: start=false
Jan 19 09:00:46 host sandmann[1574]: Appending sig: ua(yv) data: 1846, [[1, [o, /org/freedesktop/login1]], [2, [s, org.freedesktop.login1.Manager]], [3, [s, PrepareForSleep]], [8, [g, b]]]
Jan 19 09:00:46 host sandmann[1574]: Appending sig: yyyy data: 0, 4, 0, 1
Jan 19 09:00:46 host sandmann[1574]: Creating message with serial 1846
Jan 19 09:01:10 host java[1574]:         at java.base/java.lang.Thread.run(Thread.java:840)
Jan 19 09:01:10 host java[1574]:         at java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:635)
Jan 19 09:01:10 host java[1574]:         at java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1136)
Jan 19 09:01:10 host java[1574]:         at org.freedesktop.dbus.connections.base.ConnectionMessageHandler.lambda$handleMessage$0(ConnectionMessageHandler.java:91)
Jan 19 09:01:10 host java[1574]:         at eu.geekplace.sandmann.Logind.$anonfun$6(Logind.scala:194)
Jan 19 09:01:10 host java[1574]:         at eu.geekplace.sandmann.Logind.$anonfun$5(Logind.scala:172)
Jan 19 09:01:10 host java[1574]:         at jdk.proxy2/jdk.proxy2.$Proxy36.Inhibit(Unknown Source)
Jan 19 09:01:10 host java[1574]:         at org.freedesktop.dbus.RemoteInvocationHandler.invoke(RemoteInvocationHandler.java:97)
Jan 19 09:01:10 host java[1574]:         at org.freedesktop.dbus.RemoteInvocationHandler.executeRemoteMethod(RemoteInvocationHandler.java:153)
Jan 19 09:01:10 host java[1574]:         at org.freedesktop.dbus.RemoteInvocationHandler.executeRemoteMethod(RemoteInvocationHandler.java:251)
Jan 19 09:01:10 host java[1574]:         at org.freedesktop.dbus.RemoteInvocationHandler.executeRemoteMethod(RemoteInvocationHandler.java:234)
Jan 19 09:01:10 host java[1574]: Exception in thread "DBus-Signal-Receiver-1" org.freedesktop.dbus.errors.NoReply: No reply within specified time
Jan 18 21:45:17 host sandmann[1574]: Received sleep signal:
start=true

Some idle check may return a idle since period starting with the unix
epoch (e.g., [1]). Convert those timestamps to the system boot
timestamp, because that is when the idle probably started.

Alternative, we could cap the idle since start time to the system boot
time, but let's try this first.

1: https://github.com/systemd/systemd/issues/35163

As the code comment states, this is all that is required. But we still
lists the sessions if debug is enabled.

This avoids an NoSuchElementException:

Oct 16 16:02:46 neo-pc.sch systemd[1]: sandmann.service: Main process exited, code=exited, status=1/FAILURE
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann.main(Sandmann.scala)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann$.main(Sandmann.scala:75)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.util.Using$.resource(Using.scala:261)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.runtime.function.JProcedure1.apply(JProcedure1.java:10)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.runtime.function.JProcedure1.apply(JProcedure1.java:15)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann$.main$$anonfun$1(Sandmann.scala:74)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.util.Using$.resource(Using.scala:261)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.runtime.function.JProcedure1.apply(JProcedure1.java:10)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.runtime.function.JProcedure1.apply(JProcedure1.java:15)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann$.main$$anonfun$1$$anonfun$1(Sandmann.scala:74)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann.run(Sandmann.scala:164)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann.performCheck(Sandmann.scala:567)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Sandmann.idleAction(Sandmann.scala:421)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Logind.getUserSessionCount(Logind.scala:85)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Logind.getSessionProperties(Logind.scala:97)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.collection.mutable.ArrayBuffer.map(ArrayBuffer.scala:43)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.collection.StrictOptimizedIterableOps.map$(StrictOptimizedIterableOps.scala:87)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.collection.StrictOptimizedIterableOps.map(StrictOptimizedIterableOps.scala:100)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Logind.getSessionProperties$$anonfun$2(Logind.scala:97)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Logind$Session.<init>(Logind.scala:65)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at eu.geekplace.sandmann.Logind$SessionClass$.apply(Logind.scala:54)
Oct 16 16:02:46 neo-pc.sch java[110666]:         at scala.collection.immutable.Map$Map3.apply(Map.scala:397)

Only on protected branches SSH_DEPLOY_KEY is available.

Just resetting 'idleSince' to 'None' after we resume from
sleep/hibernation is not sufficient to prevent Sandmann from issuing
the idle action right away, because logind/systemd will report an
"old" idle since value of the user sessions.

To fix this, we have to introduce a barrier that prevents Sandmann
from assuming that the system is longer idle than the instant the
system resumed. This commit adds such a barrier with the name
lastWakeup.